A comparative performance evaluation of intrusion detection based on neural network and PCA

Abstract

Security is the biggest challenge for the digital data of information systems and computer networks. Some systems are used for providing security to this data. Like these systems intrusion detection system (IDS) is used for providing security to computer networks and information systems. In IDS many systems uses number of techniques for providing accuracy by selecting complete features of dataset but they lagged in terms of time and memory. For real time applications time and memory is critical issue. So, there is a need of such systems which will minimize time and memory parameters. This paper presents IDS using two Methods. These both methods based on neural network. First method uses less features of dataset using Principal component Analysis (PCA) technique and second method uses complete features of dataset. Experiments are performed on these two methods using KDD Cup 99 dataset. The results simulate the effect of less featured based incomplete learning technique and complete feature based learning technique. According to the obtained results when the system usage the less features of KDD Cup 99 dataset with incomplete instances of data then the classification accuracy of model becomes less efficient as compared to the entire dataset training but it is efficient for time and memory parameters. So, Method I is beneficial for real time applications. These both systems are developed using Java technology.